TheophilusBible is for reading Scripture closely — in English or in the Hebrew and Greek — and for turning what you find into teaching. You do not need an account to read, mark or lay out a passage; you need one to keep projects and to use the AI, which is paid for with credits and always shows its price first.

Finding your way. The menu on the left is grouped by what you are doing: Explore & Read (the text, its structures and its words), Ministry Preparation (lessons, sermons, devotionals, evaluating a sermon), Academic Research (projects, your commentary, the research graph) and Account & System. Rest the pointer on any menu item (or move to it with the Tab key) and a bubble says what that tab is for. On a phone the menu is behind the ☰ button, and every tab has its own entry below.

On a phone. A bar along the bottom of the screen holds the five places you come back to most — Home, Today (the daily devotional), Library, Search and You (your profile and saved work) — and lights the one you are in. Everything else, Structures and the Prepare and Research tabs among it, is in the ☰ menu. In the Library, the reader's display switches (word gloss, transliteration…) sit in one row you can swipe sideways, and each measured figure above a chapter shows its title until you tap it open.

A good first hour — from a passage to something you can teach:

1. Open a passage. In Ask & Study, type Ruth 1. It opens on the board. Tell Theophilus what you are preparing — a sermon, a small-group lesson, a devotion — and he will point you the right way.
2. Read it slowly and mark it. Choose an entry from the legend (God, a place, a repeated word…) and drag across words. Use Lay out to put each clause on its own line.
3. Let the data show you more, free. The buttons under the chat ask the board for a chiasm, lay the text out, light repeated words or words that share a field of meaning, mark textual variants (New Testament), or read the passage aloud. Everything Theophilus adds is labelled with where it came from and undone in one tap.
4. Write what you see. In the guide beside the text, answer the three questions for a verse at a time: what does it say (Observe), what did it mean (Interpret), what does it ask of me (Apply).
5. Find it yourself first. Guide me is on to begin with: Theophilus asks what you notice before he shows anything, then compares. Press it to turn it off.
6. Ask when you are stuck. A question in the chat is answered about the passage on the board, from this site's own data, with the price shown before the first one.
7. Make something. Make… at the top of Ask & Study turns your study into a sermon, a lesson, a devotional, a video script or slides. It is built on your interpretation, so write at least one first.

Studying with others. Share or send (under More in a study) can make a read-only link, or, under Study it together, put friends on the same study with you as editors or viewers, each person's work named.

Other ways in. Home answers a quick question and sends anything bigger to the right tool. Library reads any chapter with glosses and audio. The Prepare tabs go straight from a topic or passage to a finished piece. Research is for long projects with evidence and sources.

Your language. Pick the interface language at the top of the menu. The AI can write in your language too; choose it where a tab offers Output language.

Stuck, or found something wrong? Contact us in the menu reaches a person, and the round button in the corner opens a guide that can answer questions about the site itself.

One page: a chat with Theophilus on top, and Guided Study's board beneath it. Type a passage — Ruth 1, Malachi 3:6–12, 1 Cor 13 — and it opens on the board, where you can mark words, lay out the lines, connect and draw exactly as in Guided Study. Opening, reading, marking, laying out and the data cards are free, and need no account: without one, your board is kept in this browser.

Asking costs credits, and you see the price first. The first time you ask a question, a card says roughly what a question costs before anything is sent. Without an account it asks you to sign in; if you have no free requests or credits left, it offers to buy some. Either way your question is kept, and when you come back — from signing in, or from the payment page — it is sent for you. With a passage on the board, the question is asked about that passage and the answer is saved in your study.

Signing in keeps your work. Everything you marked before signing in moves into your account at sign-in, checked mark by mark; if any could not be kept, the chat says how many. Sharing, sending, slides and groups need an account, so they appear once you sign in. Your credit balance is shown at the top of the page.

Ask the board to do things. Type Is there a chiasm here?, lay it out, highlight the repeated words, connect the repeated words or switch to the BSB. These are free: the structure detector and the word counts answer, not the AI. A structure is drawn only when the detector found one in your passage, with its score; otherwise Theophilus says it found none. Everything he adds is marked Theophilus in Your marks (hover for why), and one tap of Undo in the chat takes it all back. If you already laid the passage out yourself, he asks before replacing it, and Undo gives your layout back exactly.

Also free: show the semantic domains lights words that share a field of meaning (in the original, or the BSB, which is aligned to it); group 1:3-5 as Loss brackets those verses under your label; note on 1:1: the famine is why they left writes your note on the verse (or note: … for the verse the guide is on). Anything else you ask the board to do — colour the words about going and returning — is read by the AI, priced like a question, and the AI may only pick one of these same actions, with the same rules and the same Undo. On a phone the chat is the first tab of the panel at the bottom of the screen.

The buttons under the chat do the same free things without typing, in whatever language the site is in: Is there a chiasm?, Lay it out, Repeated words, Fields of meaning, Textual variants, Old Testament in the New and Listen.

• A whole book: type just its name (Ruth, the book of Romans) to see its chapters and Bullinger's outline. Choose a chapter or a section to open it on the board; a tick marks a chapter you have studied, and choosing it reopens that study, so a book studied over several weeks stays together.
• Old Testament in the New: a note on each verse that is cross-referenced to the other testament — the Old Testament behind a New Testament verse, or where the New takes up an Old one — with buttons in the chat that open each on the board (Malachi 3:1 → Mark 1:2). They come from OpenBible.info's cross-references, ranked by readers' votes; not every one is a quotation, and the note says so. You can also type where is this quoted?.
• Textual variants (New Testament): a wavy red underline where the printed Greek editions differ, with a note on the verse saying what each one prints — the SBL Greek New Testament, Westcott–Hort, Tregelles, Nestle–Aland 28 and the Robinson–Pierpont Byzantine text. These are editions, not manuscripts; places where only the Byzantine edition differs are counted rather than drawn. You can also type show the textual variants.
• Listen reads the passage aloud from the same recordings Library plays, naming the voice; a synthetic one is said to be synthetic. Press it again to stop. Not every passage is recorded yet.
• Verse numbers. Where English numbers verses differently from the Hebrew or Greek — Malachi 4 is Hebrew Malachi 3:19–24, and a psalm's title is the Hebrew's verse 1 — Theophilus says so when the passage opens, so the original on the board does not seem to lose verses.

Guide me. Guide me, at the top of the page, is on until you turn it off: before Theophilus shows a structure, the repeated words or the fields of meaning, he first asks what you notice. Type your answer: it is saved as your own observation, then the board shows what the data finds and sets it beside what you said — which verse pairs you also saw, which repeated words you spotted. Just show me skips the question for that one time; press Guide me to turn it off (it stays off on this device until you turn it on again). No AI is used for any of this.

Choosing A small-group lesson also offers Invite your group, which opens Study it together for the study on the board (signed in), so the group can mark and comment on it with you.

What are you preparing? The welcome asks: a sermon, a small-group lesson, a devotion, or just exploring. Your answer is remembered on this device and puts that output first in Make….

Make… (signed in) turns your study into something you can use. It makes the study a Research project whose stages are your own observations, interpretations and applications — nothing Theophilus or the AI wrote — and opens it with the engine you chose selected; press Generate there, and you see the price first. A study with no interpretation yet is refused, because a sermon is built on what the text means, not on the passage alone. A video script is written from the lesson, so make the lesson first. Slides of this study downloads the replay of your study as a deck, free.

Try: type Ruth 1:1-5, mark "Bethlehem-judah" as a Place, then type highlight the repeated words, or ask Who were the Moabites? Or type Mark 1:1-8 and press Textual variants.

The button in the bottom-right corner opens a guide you can talk to. Ask it about the text — "where does chesed appear?", "what's today's devotional?", "is there a chiasm in Jonah 2?" — or about the site itself: what's free, what needs an account, where a particular tool lives. When there's a screen that answers your question better, it offers a link; it never moves you without you clicking.

You can type or use the microphone, and the speaker button reads replies back to you. Both use your browser's own speech features, so nothing is recorded or uploaded.

Hold the space bar to talk. With the guide open and your cursor out of the text box, hold space, say your question, and let go — it sends when you release. There is no wake word and there will not be one: nothing listens without a finger on a key.

The guide knows which screen you are on and what you have selected there, so on the Sky you can ask "what is this?" or "what else shares words with this?" about the star you have open. It can fly you to a chapter — ask for Ruth 2 and it offers the trip; you take it. What it still cannot do is act for you: it has no tool to sign you in, change a setting, or alter anything you own.

Ask what a Hebrew or Greek word or grammatical term means and it looks the answer up in cached BDB (Hebrew) and Abbott-Smith (Greek) lexicon entries or the grammar glossary, and cites the source — e.g. (BDB p. 224) or (Abbott-Smith, s.v. G26) — rather than defining it from memory. If nothing is cached for that word, it says so plainly instead of guessing. Check "Allow web search for this question" below the chat log to let it search the open web (via your own saved Exa key) for that one question — results are always labeled web-sourced and less certain, never blended into a cited lexicon answer.

Try: ask "what can I do here without signing in?", or "what does chesed mean?"

What it will not do. It can look things up and point the way, and that is all. It cannot sign you in, change a setting, start or delete a project, or buy anything — those tools don't exist for it to use. It will never ask for a password, an API key, or card details, and you should never type one into it. Signing in happens on the sign-in screen, in your browser, where the guide never sees what you type.

A new entry every day, built on one feature of the Hebrew or Greek that English translation cannot easily carry — an infinitive absolute doubling a verb against itself, a Greek middle voice, a word used once in the whole canon. Each entry moves from that observation to worship and ends with one concrete thing to do.

No account needed, and nothing to generate — the day's entry is already written. Open "The words behind this" at the foot of an entry to see the exact words and morphology the observation rests on, and click the reference to read the whole chapter in Library.

Try: read today's entry, then click Browse past devotionals to work back through the archive. Every day has its own link you can share or bookmark.

Where the entries come from: a deterministic sweep over this site's tagged corpus finds verses that genuinely carry a checkable feature, and the writing is built around that specific finding rather than a free-ranging search for something interesting. Entries are reviewed and approved by a human before anyone sees them.

They also wait at the verse. Every published entry appears in Commentary beneath the verse it was written about, collapsed, however long ago it was published — so a devotional you never saw on its day is still there when you come to the passage. Untick Devotional in the layer toggles to hide them.

One box, no need to pick a tab first — type anything and the Service figures out where it belongs. A bare topical or theological question ("what about predestination?", "were there dinosaurs in the Bible?") gets a quick answer right there on Home, with the most directly relevant passages listed underneath — nothing is saved and no project is created. A question anchored to a specific passage goes to Read; wanting to build or start something goes to Research or a Prepare wizard.

Try: type "what about predestination?" and watch the answer appear inline, disputed views named on both sides.

If the question names a specific Hebrew/Greek word or passage, the answer is grounded in cached lexicon and corpus data first — a "Grounded in: ..." line lists what it cited, and it says so plainly rather than guessing when nothing is cached. The checkbox next to the provider dropdown opts that one question into an open-web search too (via your own saved Exa key), always shown as a separate, clearly-labeled "less certain" section.

The greyed-out example text in the box is a real suggestion, not just a hint — with the box empty, press the right-arrow key to fill it straight in (same idea as a terminal's command autosuggestion), then edit or submit as usual.

When you are signed in, a Recent work panel appears above the box listing what you have generated lately. Generation does not stop when you close the tab — an Exegesis or Read request finishes on the server whether or not the page is still open — so this is where to look for something you walked away from. Anything still running says so, and you can safely close the tab again. Finished Exegesis and Read results open in place; drafts that belong to a project or a sermon send you to that tab, where they were saved.

To clear the list, tick the boxes beside the entries you are finished with and use Remove Selected, or Remove All to empty it. Removing an entry removes the record of the job, not anything you saved from it — a draft already sent to your Library or a project stays exactly where it is.

Try: start an Exegesis, close the tab, then come back to Home a minute later.

Every sentence of the Hebrew Bible and the Greek New Testament, diagrammed. Pick a book and chapter, then choose a depth:

1. Flow puts each clause on its own line, indented under the clause it depends on, with its joining word ("that", "because", "in order that") in front and a label saying what it does. Read down the left edge for the main line of thought.
2. Clause marks each part by its role — subject, verb, object, complement, adverbial — and reads out each clause's kernel: who does what.
3. Line draws one sentence as a Reed–Kellogg diagram: the kernel on a baseline, every other word hanging beneath the word it modifies. Hebrew is drawn right to left. Download SVG saves it for a handout or slide.

A diagram is drawn for a whole sentence, not a verse, so a chapter can begin with a sentence that started in the chapter before. Hover any word to see it in every depth, with its gloss, lemma and parsing.

The syntax trees are the MACULA Hebrew and Greek datasets (CC BY 4.0). A few things are ours rather than theirs, and each is listed under Where these lines come from: names such as "purpose" or "reason" are suggestions from the joining word, not findings, and a word the sentence leaves unstated is shown as (X), never supplied.

Read a passage slowly and mark what you notice, the way inductive Bible study marks a printed page. Start a study by choosing a passage and whether to read it in English (Rotherham's Emphasized Bible, the Berean Standard Bible, Young's Literal Translation or the King James Version, all free to use) or in the original Hebrew or Greek. In the original you choose how much help to see: the text alone, or with transliteration, an English gloss under each word, and parsing. You can switch texts at any time; marks made in each text are kept separately.

Marking. Pick an entry from the legend (God, a key or repeated word, a contrast, a time, a place, a command, a promise…), then drag across the words with a mouse or pencil. On a phone or tablet, tap the first word and then the last. Every entry can be renamed and restyled, and you can add your own. Format freely gives any mix of highlight, bold, underline (single, double, wavy or dashed), box, circle and symbol in any colour. Add a note attaches a short note to the words you select. With a keyboard, the arrow keys move through the words and Enter marks.

Colour. In the Hebrew or Greek, Colour paints each word by its part of speech, or colours just the verbs: in Hebrew by where they stand in the storyline (main line, secondary, background, setting), in Greek by how prominent their tense form makes them. The colours are the ones the Exegesis tab uses, and a small key under the bar names the ones on screen. The storyline comes from automated tagging, and the key says so. Your marks draw over the colours, so you can use both at once. Reading the Berean Standard Bible, Colour paints the English too: each word takes the colour of the Hebrew or Greek word it translates, as matched word by word by people at Clear Bible — a word nobody matched stays plain. Original beside sets the Hebrew or Greek under each verse; point at an English word, or tap a word of the original, to see which words translate which.

Saved as you go. Guided Study needs an account: each mark is saved to it the moment you make it, in order, so a study is never lost with a browser's cache and opens where you left off on any device. The Your marks tab lists your marks in the order you made them: tap one's reference to see its words lit up in the passage, remove any of them, or undo the last (Ctrl+Z, or ⌘Z on a Mac). No AI is used, so studying costs no credits.

Room to read. The passage takes the page; the controls sit in one bar above it. Mark, Lay out and Draw switch what a tap or a drag does, and the legend shows only while marking. A− and A+ set the size of the text, and Focus puts the site menu away while you study; both are remembered on this device. On a wide screen the guide sits beside the passage; on a phone or a tablet held upright it waits in a tab at the bottom of the screen — tap Guide to open it and tap again to put it away. Downloading slides, other versions and deleting the study are under More. So is Print a handout: the passage as it stands on the page — laid out if you laid it out, with the layers you have on — and ruled room to write under every line, for a group or a class. Choose Save as PDF in the print dialog for a file.

Guide me verse by verse. The guide takes one verse at a time and asks what you notice; what you write is saved against that verse. Want more? opens the next layer of what the data holds, free: the words (gloss, transliteration, dictionary form, parsing), the grammar of the verbs, the clauses, how they connect (main line or background, the relation and the connecting word), and the words the passage repeats. Word study beside a word opens its dictionary entry, how often it stands in the passage, the book and the Bible (and in which verses of this book), and the other words UBS files in the same field of meaning — free. How the Septuagint renders a Hebrew word is not shown yet, and it says so. I'm tired, just give me the information opens every layer at once. The discourse layer comes from automated tagging and says so. The two buttons marked AI credits — a response to what you noticed, and a written summary of the verse or passage — are paid, and each answer is saved into your study. So is Ask a question about this passage, folded away at the foot of the guide.

Observe, interpret, apply. The guide follows the three questions of inductive study. Observe asks what you notice; Interpret asks what the verse meant to its first hearers; Apply asks what it asks of you. Each is saved against the verse under its own name, and the paid response reads it for that stage — testing an interpretation against the grammar and context, or keeping an application to what the passage itself says.

Other languages. Besides English, a study can be read and marked in the Reina-Valera 1909 (Spanish), the Louis Segond 1910 (French), the Luther Bible 1912 (German), the Statenvertaling (Dutch), Ang Biblia 1905 (Tagalog) and the Chinese Union Version (1919, simplified characters) — all free to use. Choose them from Other languages, next to the English texts. Chinese is written without spaces, so its words were divided once, when the text was added; a division can occasionally be wrong, but it never changes, so your marks stay where you put them.

Other versions. Copyrighted translations such as the NIV or ESV cannot be copied into the app. Read this passage in another version, under More, opens the verse you are on in YouVersion or Blue Letter Bible, where their publishers license them; your marks stay here, on the texts the study can show.

Lay out the lines. Set the passage out the way a poem or a sequenced list is printed. Choose a phrase — drag across it, or tap its first word and then its last — and press Start a new line here to bring it down to a line of its own start, or Put it on its own line to break the line after it too. You can also tap between two words to start a new line there, and the ↑ beside a line joins it back to the one above. Tap lines in their left margin to select them, then indent them, tag how each relates to what it hangs on (purpose, reason, result, contrast, series, parallel…), or bracket them into a labelled group. Compare shows how the Hebrew or Greek itself divides and indents the same verses, with the same relation names, so you can read your layout against it. Below that it shows Bullinger's structure from The Companion Bible (1909–1922): the lettered outline he set the passage in, with the part your passage falls in highlighted. Most outlines have not yet been checked against the printed book, and some letters were supplied by the editors rather than printed by Bullinger; Compare says so beside each one.

New to diagramming? In Lay out, Start with a default diagram sets the passage out for you as a first draft. Reading the Hebrew or Greek, it follows the clause analysis the Diagram tab draws; in a translation, it starts a line at each full stop, semicolon or colon, at a comma once a line has some length, and before words such as because, so that or although, indenting and tagging those clauses (reason, purpose, concession…). It is only a starting point: move, indent, retag or join any line. It never replaces a layout you have made without asking first, and Clear the layout takes every line break, indent, tag and bracket away at once while leaving your marks, notes and connectors.

Connect words and phrases. Connect draws a line from one word to another: drag from the first to the second, or tap one and then the other. Set Join to Phrase to phrase to connect whole phrases instead: choose the first phrase (drag across it, or tap its first and last word), then the second, and both are bracketed beneath their words. Choose beforehand what the line says — a relation such as purpose, reason or contrast, or a label of your own — and its colour, style and arrow. Every connector is listed under Connectors in this passage, where each can be given a different relation after it is drawn, or removed. Connectors stay with their words when you lay the passage out, change the text size or open the study on another device. In Lay out, choose lines and the line they hang on and press Attach: they are indented under it and joined to it, the way the Diagram tab draws one clause under another. Reading in the original, Compare offers Start my layout from this, which sets out your lines where the Hebrew or Greek's own clauses begin, ready for you to change. Slides list your connectors on a slide of their own.

Ask a question. The guide also takes your own questions about the Bible, theology and church history, answered with the verse in view (AI credits). Answers start from the historic creeds most Christians share (the Apostles' and Nicene Creeds, with the Chalcedonian Definition of Christ as one person, fully God and fully man) and describe other traditions fairly; say how sure the history is; name the passages that bear on the question; and keep to their subject, pointing you elsewhere for anything else. If a question shows you are in distress, the answer will point you to your church and, where there is risk of harm, to emergency help. Every answer is saved in your study.

Apple Pencil and stylus. On an iPad, a foldable phone or any screen with a pen, stroke across words with the pencil to mark them with the chosen legend entry; the first pencil stroke switches the passage into pencil mode, so strokes mark rather than scroll, and a palm resting on the screen is ignored. Draw switches to ink: draw arrows between words, brackets across lines, or notes in the margin. Each stroke is saved as you lift the pencil and moves with the word it starts on when you lay the passage out differently.

Study with a group. Open the Group tab and choose Study this with a group, then send the link to the people you are studying with. Each person who opens it (signed in) gets their own copy of the study. As each one saves what they noticed, what it meant or what it asks of them, the Group tab shows it for that verse, so you can compare. Your marks, notes, lines and drawings stay yours — the group never sees them. The leader can make a new link (the old one stops working), remove someone or end the group; everyone keeps their own study. People are shown by their account name, never their email.

Download slides. Turns your study into a PowerPoint deck for teaching: the text, then one slide for each thing you marked with every mark so far on the words, your laid-out lines, what you noticed, and the legend. What you noticed and any AI responses go into the speaker notes. PowerPoint cannot draw a box or circle around words, so those appear as coloured brackets. If you drew with the pencil or connected words, an As drawn slide shows your strokes and connectors on the words they belong to. To make a PDF, open the deck in PowerPoint, Keynote or Google Slides and save it as PDF.

Share, send or delete. Share or send — beside each study in My studies, or under More inside one — makes a link to the study. Anyone who opens it can read the study, laid out as you left it, with its marks, notes and saved answers, without signing in; they cannot change it, and signed in they can Save a copy to my studies and carry on in a copy of their own. Copy the link, email it, or send it with another app; on a phone or tablet you can send the slides themselves. Stop sharing makes the link stop working for everyone. Delete study, in the list or under More, deletes a study and its link for good.

Study it together. Further down the same Share or send dialog, Study it together puts other people on this study rather than giving them a copy. Invite someone by their email as an editor (they mark, lay out and ask on the study with you) or a viewer (they read it and comment). The invite link is copied for you to send, and works only for the account with that email. To let anyone signed in join with a link instead, choose Anyone signed in with the link; switching back to Only people I invite stops every such link at once.

On a shared board every mark, note and answer is named with who made it, and what someone else adds appears on your page within a few seconds, without reloading. An editor can remove only their own work; you, the owner, can remove anything, take someone off the board, withdraw an invite, and hide a comment. Each person keeps their own view (which text, how much help) and pays for their own AI questions. What other people write is never treated as an instruction to the AI: it reaches it only marked as someone else's words, and comments never reach it at all. Make… uses only your own observations and interpretations. A board someone shared with you is listed in My studies as theirs, with Leave this board.

Layers and Replay. Layers, in the toolbar, chooses what the page draws: marks, notes and answers, lines and structure, connections, ink, and Theophilus's work — and on a shared board, each person. It only hides; nothing is deleted, and your choice is kept on this device. Replay shows the study as it stood after any step: drag the slider, or press Play to watch it build up in the order it was made (the same order the slides use). While replaying you cannot mark the page; Back to now returns to the study as it is. Both work on a shared link too. Commentary (sticky notes), off until you turn it on, adds the Commentary tab's notes on these verses as yellow notes under each verse, each with its confidence (high, moderate, low, disputed). They are written by AI, say so, and are never saved into your study, so they cannot be mistaken for what you noticed. Only commentary already written is shown; nothing is generated or charged. Discourse, also off until you turn it on, shows how the account is built: words off the main line (setting, background, comment) are lighter and slanted, a peak is boxed, and marked word order is overlined. It reads the corpus's own annotation, most of it automated and said to be so; in a translation it works through the BSB's alignment.

Try: start a study of Ruth 1:1–5 in Rotherham, choose "Place" in the legend and drag across "Bethlehem-judah". Then choose Lay out and Start with a default diagram.

Curated comparisons: the tabernacle beside Solomon's house and the house Ezekiel is shown, drawn at one scale; two words that occur together in three verses of the whole Hebrew Bible; Exodus 12 beside the death of Jesus.

Every cell cites its verse, and each entry says who makes the connection — that is the part to read first. A measured entry compares numbers the texts state, and marks each one stated, derived (with the arithmetic printed) or unstated. A lexical entry is a word the passages share, with its count in the annotated corpus rather than the word "rare". An explicit entry is one a later verse makes, and that verse is printed on the arrow. An interpretive entry is a reading, named and labelled, and it never draws an arrow, because no verse makes the link.

Where the sources disagree — Kings against Chronicles on the porch's height, John against the Synoptics on the night of the Passover — both readings are shown and neither is resolved. Comparisons are free to read, and so is every verse they link to in Library.

Entries also appear beneath the passage itself in Library and Commentary, collapsed, with a line saying why that chapter summoned them.

Browse and read the Hebrew/Greek text with English glosses.

1. Click a book card.
2. Pick a chapter from the chapter navigation.
3. Toggle "transliteration" if you want it shown alongside the original text.

Try: click Ruth, then chapter 1.

Click any Hebrew or Greek word for a popover with its lemma, gloss, and part of speech. Toggle flowing to read a chapter as continuous text with small verse numbers instead of one card per verse — handy for long chapters like Psalm 119. Toggle fluent translation (KJV) for a normal-English reading alongside the original; a small number of chapters (about 1 in 8, mostly in the Psalms and a few other Old Testament books) show a note instead, because the Hebrew and King James verse numbering genuinely differ there and showing mismatched text would be worse than not showing it. The current book/chapter is also reflected in the page URL, so you can bookmark or share a link straight to a passage.

explore.theophilusbible.com is a second site, and a different thing from this one. It is a reader: Rotherham's Emphasized Bible (1902) laid out on the structures E. W. Bullinger drew for The Companion Bible, with the seam between one structural member and the next ruled through the text. No account needed, the whole Bible — as with this Library.

This Library is the other half: the annotated Hebrew and Greek corpus, with morphology, the syntax layout, search and commentary. The distinction is not English against the original languages — Explore shows the Hebrew and Greek too, from this very corpus, coloured by the same palette. The distinction is what you came to do. Read a book through, over there; ask a question of the text, here.

The Hebrew and Greek over there are free too, with no account. Click Original in Explore's toolbar to show them beneath the English. Rotherham and Bullinger are public domain; the Hebrew is derived from the ETCBC's BHSA, whose licence is non-commercial, so Explore shows it free for reading and credits it on the page, as this Library does on the Sources page.

The two are linked at the chapter. Read on Explore beside a chapter title in Library opens that same chapter in the reader, and Study in the reader's toolbar brings you back to it here. Explore also underlines the words a Daily Devotional touches and links back to the devotional on this site.

A Section Through Sacred History is one chart read left to right, from Abram to the present. Five strata run its whole length: what Scripture records on top; beneath it the doctrines the church has argued over, each with the verses both sides argued from; then the kingdoms and rulers whose paths cross the biblical story; then the wider world's writing, thought and material remains; and a bedrock strip tracing peoples and empires as continuous threads.

1. Open Timeline in the sidebar. It is its own page and opens in a new tab, so anything running here keeps running. Back to TheophilusBible at the top of the chart returns you here, still signed in.
2. Jump with the era bar, or search for a person, place or text.
3. Click any entry for its photograph, its fuller note, and its links into Library and Commentary on this site.

The controls above the chart change what is on screen. Reading view puts the chart away and gives you the same entries as a plain outline to scroll top to bottom. Connections draws the dashed curves that tie a find to the passage it bears on. Pictures and Daylight turn the photographs off and the dark ground light. The zoom control widens or narrows the whole span.

The chart's own Key says how it settled its dates, and it is worth reading before arguing with one. Dates before the monarchy are approximate and contested; the regnal years of Israel and Judah follow Thiele's synchronisms; the exodus is plotted at the early date implied by 1 Kings 6:1, with the thirteenth-century alternative noted on the card itself; Genesis 1–11 opens the chart as a prologue, outside the measured scale; and a tilde marks any date scholarship still disputes. Distance along the chart is not time — each era is given room according to how much happens in it — so read the date on a card rather than the gap between two of them.

Try: search for Siloam, then follow the dashed curve from the inscription to Hezekiah's tunnel and open 2 Kings 20:20 from the card.

An indented view of a chapter that shows how its clauses hang together — each clause set at a depth matching how it depends on the one above it, rather than laid out one verse per line.

1. Open a chapter in Library.
2. Tick Syntax Layout.
3. Hover any relation badge to see what it means, or open the legend card below the text for all twelve at once.

The badges are the corpus's own clause-relation codes — Objc for a clause acting as an object, Coor for one joined by a conjunction, Adju for an adverbial modifier, and so on. Clauses whose relation the corpus does not record show no badge at all rather than a placeholder, so a gap means "not annotated", not "no relation".

This layout is read off the corpus annotation, not generated: no model is involved, and no interpretation of the passage has been applied to it. Like every annotation layer here it is open to correction — treat an odd-looking depth as a lead to check against the text, not as a verdict.

Try: Ruth 1 with Syntax Layout on, and follow the indentation through verses 8–18.

An original-language commentary organised by pericope — the interpretive unit — rather than verse by verse. The Hebrew and Greek, the morphology and the colour coding are the same fixed corpus data Library and Exegesis read; only the prose around them is written.

1. Pick a book. Filter to Old Testament or New Testament if the grid is long.
2. Pick a pericope from the inventory. Each row says which method drew its boundaries and how much confidence that method earns.
3. Read. Use ← Previous and Next → to work through a book in order.

Try: Ruth, then the first pericope. Every book's commentary is free to read, with no account needed.

Where the units come from, and why they carry a confidence. A pericope boundary is somebody's interpretive decision, not a fact about the text, so the app never presents one as settled. Divisions are drawn deterministically from the corpus's own paragraph markers where those are the right grain, and from chapter units where they are not — the label on each unit says which. Open "How else could this book be divided?" under the inventory to see the alternatives, with a count of how many detected literary structures each one cuts across. A structure running through a boundary is evidence that the boundary is in the wrong place.

Every interpretive claim shows its grounds. Claims carry a confidence label, and Why? beside one opens the textual evidence it rests on and any alternative readings the writer considered. A claim that was given no supporting evidence says so rather than staying quiet about it. If you disagree, use I disagree or Not convinced — your stance and optional note sit beside the claim and never alter the prose. Checked marks one you have satisfied yourself about.

Layers at the top of a pericope turn parts of the reading on and off: Text, Translation, Structure, Vocabulary, Grammar, Modern Hebrew, Discourse, Textual and Commentary, plus Devotional for any daily devotional written on a verse in the passage. Turn off what you are not using and the page gets shorter.

The measured band above the verses is counted, not written — literary structures found in the passage, discourse analytics tabulated from the corpus's own annotation, and any figures the chapter summons. Each panel opens onto its working. Read it against the prose below rather than as a summary of it.

Click any Hebrew or Greek word for its lemma, gloss and morphology, the same popover Library gives you. Where the manuscripts disagree, the variant is marked in the text and the reading is set out with its sigla. Recorded passages get a play button per verse. Open in Library, Exegesis or Structures carries the same passage to those tabs.

Generating and correcting. A pericope with no commentary yet shows the corpus text and a Generate commentary button; the text itself is always there whether or not anything has been written about it. If one verse reads badly, Rewrite this verse redoes that verse alone rather than the whole unit — and you can say what should be different. Book-level orientation is generated separately from the book page.

Commentary prose is interpretation and is labelled as such at the top of every pericope, with the model and date that produced it. Weigh it as you would any commentary. The Hebrew and Greek are never generated.

On a Hebrew pericope, the measured band carries a collapsible panel listing the clauses where a subject-like word stands before the verb — and, where it has been generated, how the same thing would ordinarily be said in Modern Hebrew.

The modern sentence is not the point. The difference is. Hebrew narrative normally puts the verb first. When something else stands there instead, you cannot see that it is an unusual position without knowing the usual one — and Biblical Hebrew gives you no such baseline from inside itself, because every sentence in it is old. Modern Hebrew does. So the modern line is a control: read the two, see which word has moved.

1. Open a pericope in any Hebrew book in Commentary.
2. Make sure the Modern Hebrew layer is ticked.
3. Open How it would be said today in the measured band.

Two halves, and they are not the same kind of thing. Which clauses are subject-first, and which word stands first, is computed from the words themselves — so it is there on every Hebrew pericope whether or not anything has been written about the passage. The Modern Hebrew rendering is generated, and carries a confidence and the model that produced it. A clause showing the computed half but no rendering simply has not been generated yet — a fact about this site, not a finding about the Hebrew.

You never have to generate anything, and you are never charged for this. The renderings are produced once for the whole corpus and then read by everyone.

Why most verses have nothing. The ordinary backbone of Hebrew narrative — the long runs of wayyiqtol — is verb-first, so it never appears here. Neither do verbless or nominal clauses, which have no verb to stand before and therefore no order to observe. On a Greek pericope the panel never appears at all. Expect a handful of clauses in a chapter, not one per verse.

What it deliberately will not tell you. It says what moved. It does not say the word is emphatic, or that the order carries focus, topic or contrast. That restraint is not timidity: this corpus's own word-order annotation was tested against the text and could not support those conclusions, so asserting them here would be borrowing certainty that does not exist. The two lines are shown side by side and the inference is left to you — if you read Hebrew, you are better placed to draw it than the annotation is.

Word order in Biblical Hebrew is shaped by genre, clause type and particles as well as by information structure, so a word standing first is not automatically significant. The panel is a claim about Hebrew, not about theology. A sermon point that needs the word order to stand up does not yet stand up.

In Prepare a Sermon. The strongest of these travel into sermon preparation as grounding, so a point can show a hearer where the Hebrew puts a word. Only high or moderate confidence entries cross over, no more than four per passage, and the same restraint applies: show the difference, do not claim what it proves.

Try: Ruth, first pericope, with the Modern Hebrew layer on.

Find every verse using a particular word or phrase.

1. Type a search term.
2. Scan the matching verses.
3. Click a result to jump to it in Read.

Try: search chesed to see every verse using that Hebrew word.

A one-shot, plain-language explanation of a passage — no project or setup, just a quick answer.

1. Pick a passage.
2. Optionally type a specific question.
3. Click "Explain this."

Try: Ruth 1, question "Why does Naomi change her name?"

Build a personal Hebrew/Aramaic/Greek flashcard deck from words you actually meet in your own reading, with spaced-repetition review (sign-in required).

1. Pick a passage and click "Show words" to see its tokens.
2. Click a word to add it to your deck.
3. In "Review due cards," each card shows the word with the verse you met it in, the word marked. Reveal the answer and rate yourself Again/Hard/Good/Easy — the schedule adjusts automatically. Beside it are your deck's counts and the cards coming up next.

Try: show words for Ruth 1:1, add "chesed" if it appears, then review it once you have a due card.

The whole corpus drawn as one picture: the canon as an orrery of chapters, the vocabulary as two galaxies of words, and your own claims ringed over the chapters they anchor to. Every mark is a row already in the store — a star's size is its token count, its brightness the share of its annotated tokens on the storyline mainline, and a thread is rare vocabulary two chapters share (the same markers the Bible Genome uses). Nothing is generated and nothing is simulated: the same corpus always draws the same sky.

1. Open the Sky tab. The first visit builds it from the corpus, which takes a moment; later visits are instant, and "Rebuild" redraws it if the corpus has changed.
2. Click a star to see the chapter's numbers, the chapters it shares rare words with, and any claims of yours anchored there. Drag to pan, scroll to zoom, double-click to reset.
3. Switch to Vocabulary for the two galaxies of words — colour is part of speech, brightness is frequency, and a word's place is which chapters it shares with other words, not what it means.
4. Switch to Names to light where a proper noun stands. Pick a second name and the chapters holding both light in a third colour — which is the only relation between two names this corpus supports. It says where a name occurs and never who or what it is: the tagging marks a proper noun and does not say whether it is a person, a place or God.
5. Switch to Structures to light the chapters carrying a chiasm or parallelism that passed a permutation test, brightest where the test is most confident. Click one and its figure is drawn: a chiasm folds so that every mirrored pair sits on one level rung and the arms converge on the centre, a parallelism stands as a ladder with the two panels side by side, and a staircase runs one step across and one down. The shape is the pairing the detector used, so the four kinds look different at a glance. A ring on a unit is marked word order in its verse — a second measurement in the same frame, not a claim that the two are connected.
6. Switch to Word and type a word — in Hebrew or Greek, or just its English gloss — to ask where it stands. Every chapter holding it lights, brighter where it occurs more often. Pick a book from the In box and those chapters take the accent colour, with the count inside the book set against the count everywhere else. Unlike the galaxy, which draws the 2,000 commonest words per language, this finds any word in the corpus, and tells you when the one you found has no star.
7. Type a second word in the with box to ask where the two stand together. The canon then wears the Names layer's three colours: the first word's chapters, the second's, and in a third colour the chapters holding both. The panel prints how many chapters each word reaches, how many they share, what share of each word's range that is, how many verses hold both — a much tighter fact than a shared chapter — and roughly how much overlap two words this frequent would have by chance. None of that is a claim that the two words are related: two words in one chapter is co-occurrence, and the panel says so beside the numbers.
8. Under Brightness, switch the word layer between occurrences and per 1,000 words. A long chapter has more room for any word, so a high count is partly a fact about the chapter's length; the second measure divides by the chapter's own size. It says how concentrated the word is there, which is not how much it matters.
9. Open Verses on any chapter in the list to see the verses the word actually stands in, and click one to open it in the reader with that verse marked. Copy these numbers puts the whole table on your clipboard — with the caveats, which travel with the figures rather than staying behind on the screen.
10. The address bar carries the question, so a finding can be shared: a word, a second word, the book you scoped to and the measure you chose all live in the link, and so does a lit name. Paste it into a handout and it opens on the same picture.
11. In Vocabulary, use Find a word to reach one star among thousands, and Hide words commoner than to take the top of the frequency range out of the picture. Filtered stars dim rather than disappear and nothing moves: a star's place is fixed by the data, so the same corpus always draws the same galaxy.
12. "As a list" is the same data as a table, for screen readers and for checking a number.
13. 3D turns the flat chart into a sky you can orbit — drag to turn it, scroll to zoom, arrow keys to steer, double-click to reframe. It draws exactly the same rows in the same places; what it adds is depth and travel between the two galaxies. On a laptop it is on to begin with, on a phone it is off and offered; either way your choice is remembered, and a browser that cannot draw it says so and keeps the chart.

What it refuses to draw, on purpose: Greek chapters have no discourse annotation, so no Greek star is brighter than another; no thread crosses the testaments, because no row in the store links them; and every Hebrew share is labelled heuristic · unreviewed, because that is what every annotated token in the corpus still is.

Structures refuses three more things, and each is a way a picture could mislead. A chapter the detector examined and found nothing in is drawn dim; a chapter it has never scanned is drawn dimmer still, because those two mean opposite things. An even-sized chiasm's arms stop short of meeting, because an even mirror has no centre unit and drawing one to a point would invent it. And no peak is lit anywhere: this corpus carries two rival peak columns that agree on only 37% of the tokens either one flags, so neither is drawn and neither is quietly preferred.

One thing the Word layer will not do, on purpose: tell you a word is important. It counts where a word occurs and leaves the weighing to you. A word can carry a whole book while occurring three times in it — chesed occurs three times in Ruth, once each in the first three chapters — and a word can occur constantly somewhere it is doing very little. Counting is the part a computer can do honestly.

Try: open the Sky, click Ruth 2, and follow its threads to the chapters that share its rare words. Or open Word, type redeem, and set the book to Ruth.

Visual maps of a passage's discourse structure — chiasms, parallelism, repeating patterns.

1. Pick a passage.
2. View the generated structure map.
3. Click a unit to highlight its mirrored counterpart.

Try: Genesis 1 to see the creation week's repeating pattern.

Switch to Book Map to scan a whole book at once — one tile per chapter. The first scan of a book takes roughly 15–20 seconds per chapter (real detection work, not a delay), but only ever has to happen once; every result is cached, so later visits are instant.

Reading the map. A tile carries two separate facts. Its colour is what was detected in that chapter — chiasm, parallelism, staircase or chiastic parallelism — and the pattern's name is written underneath, so you never have to remember which colour is which. Its dot is how surprising that match is: filled for p < 0.01, half-filled for p < 0.05, hollow below that. The dot is three fixed states and never changes size, because a p-value is not a measure of how strong a chiasm is — it says how rarely a reshuffling of the same chapter scores as highly, which is a different question.

Two things the map cannot show you, which the legend says instead. The significance test reshuffles each window 200 times, so the smallest p it can ever report is 1/201 — a chapter sitting at that floor is bounded by it, not measured at it, and two such chapters cannot be told apart by this test at all. And a blank tile does not mean there is no structure in that chapter: it means nothing cleared p < 0.10 among the top five windows the scan looked at. An unscanned chapter is shown faded, which is a third thing again.

Upload or paste a sermon and get a structured reading of it: eight dimensions with the evidence behind each — including a Socratic/classical rhetorical-strategy read alongside the appeals triangle below — a diagram of where its argument carries and where it stops, a check of every claim it makes about the Hebrew or Greek, and a list of passages whose style suggests they may not be yours.

You will not get a score. Not a grade, not a percentage, not a ranking. These dimensions don't average — hermeneutical accuracy and vocal variety are not the same kind of thing — and a single number would invite comparison between preachers, which helps nobody. The four small blocks beside each band are not a score either: they are that band's position in an ordered list of four — absent, underdeveloped, present, strong — and nothing adds them up. A dimension marked "not assessed" shows four empty blocks rather than one filled one, because that is the absence of a judgement, not the bottom of the scale. Each dimension also says whether it is something the tool can genuinely check against the text, or an observation for you to weigh. Ethos and doxology are the second kind; a machine can see their shape but not whether they were earned.

The most useful part is usually the Spine: your sermon's load path from the text, through the truth, to the implication, the imperative, transformation, and the glory of God. Each link shows whether it carries, strains, or breaks — so you can see exactly where the argument stops holding. A break between truth and implication means an imperative with no textual parent; a break before the last step means the sermon landed on self-improvement.

Try: paste a sermon you have already preached, and fill in "Your own assessment first" before you submit. You will get to compare what you thought with what it found.

Written sermons only for now — manuscripts, outlines, or transcripts as .docx, .pdf, .txt, .md, or pasted text. Audio and video aren't supported yet, so the delivery dimension is simply not assessed rather than guessed at. Your file is read once and then discarded; only the text is kept.

On the originality check. It looks for paragraphs whose style sits well outside the rest of your own writing, and it is careful about what it doesn't flag: anything matching scripture is removed first, and so is anything with an attribution nearby. It reports passages worth checking, never a verdict — a quoted hymn, a read liturgy, a story told in someone else's voice, and your own earlier published writing all look the same from where it stands. The optional web search is off unless you tick it, and it is the only part that sends anything off this machine.

How this compares. Below the dimensions, the report checks your sermon against two things this app can itself produce for the same passage: a fresh Exegesis-tab run (what the text itself surfaces, and where your sermon agrees, adds beyond it, or misses something), and — only if you have already generated one — the sermon Prepare a Sermon would produce for the same passage, compared on main thrust, structure, and application. The second comparison is never generated on the spot, since a fresh one takes several minutes; if none exists yet, you'll see a link to Prepare a Sermon, and re-running the evaluation afterward picks it up automatically. Neither comparison is a second opinion to defer to — both are a check on the text versus what your sermon does with it, and both feed into the report's opening summary rather than sitting apart from it.

Output language. The Output language field on the form sets the language of everything the evaluation writes about your sermon — its notes, suggestions, and the "How this compares" findings. It never touches a verbatim quotation pulled from your own sermon, which always stays in whatever language you actually wrote it in. This is separate from your interface language (the dropdown in the sidebar), which controls the rubric's own fixed labels — dimension names, bands, and so on — shown around that generated content.

Plan a run of sessions on one topic or through one passage — a two-part mini-series, a summer through a book, or a full year through one. Define the series once; generating each session is a separate, deliberate step.

1. Create a series: give it a title, then either a passage (a book or a range within it) or a topic.
2. Set how many sessions you want and click Auto-divide (instant and free for a passage, snapped to real paragraph breaks where the corpus has them; one quick AI call for a topic, which asks it to suggest that many sub-topics).
3. Adjust any session before saving the plan — its title, its own portion of the passage or its own sub-topic, its style, how long it runs, and whether it should close with an altar-call emphasis. Nothing is generated yet at this point.
4. Once the plan looks right, save it, then generate sessions one at a time, whenever you are ready for each — never all at once.

Session length can vary. A camp week with a longer evening session and shorter morning ones, or a series where one session runs long and the rest are brief — each session's own duration is independent of the others.

Style, per session. Most sessions are a sermon, but any single session can instead be a workshop (the Teaching engine's leader/participant format — useful for a hands-on or discussion-style session in the middle of an otherwise preached series) or a short devotional.

Series continuity. From the second generated session onward, each one is told what the series has already covered — its theme and a short summary of every session already generated — so it builds on what came before instead of re-explaining background the series already gave, or repeating the same illustration.

Try: create a passage series on Ruth 1:1–4:22, set 4 sessions, and Auto-divide — one chapter lands naturally in each session. Generate session 1 to see it in full.

Generating a session costs the same as one full run of Prepare a Lesson/Sermon/Devotional (its own set of AI calls, a few minutes) — planning a 52-session series is instant and free, but generating all 52 would cost roughly 52 times that, which is why nothing generates until you ask for that specific session.

Pick an angle. Rather than starting from a blank question box, the Exegesis tab offers a set of ready-made focus questions — where does this passage turn, what are the verbs doing, what is being emphasised, where does the grammar strain, what does English lose here. Each one was written and tuned once, carefully, so a request buys a specific analysis instead of whatever you managed to phrase in a hurry. Pick one, type your own question, or do both: a lens sets the angle and your question adds your emphasis within it.

The list changes with the book you choose, and that is deliberate. The Hebrew Bible in this corpus carries full discourse annotation, so lenses about storyline, boundaries and fronting have real data underneath them. The Greek New Testament carries full morphology but not those layers — so those lenses are not offered there, rather than being answered from general knowledge and charged for.

Try: choose Jonah 1, pick "Where does this passage turn?", and run it. Then switch the book to Mark and watch which options disappear.

Compare two models — or three. Behind the "Compare two models…" button you can run the same passage and the same method through two different providers, or add an optional third, and read them side by side. There is no winner and no judge. Where they diverge, the text is the arbiter — not whichever model sounded more confident. This is the one place in the app that deliberately costs more: one request per model. The third column is off by default, and the line above the button states the multiplier for whatever you have currently selected.

Reading the attention band. Above the columns is a bar split into segments: verses every run cited, verses only one run cited, and — with three runs — verses some but not all of them cited. Each segment's width is simply how many references fall in it, each prints its own count, and the key underneath lists the actual verses, so you can go and check them. It used to be a single percentage, which sounded more precise than it was: "43% agreement" hides whether that is three references out of seven or three hundred out of seven hundred, and those are not the same finding. Note what the band counts — references extracted from each run's prose, not claims. Two runs can cite the same verse and say opposite things about it.

Full AI-powered discourse analysis. It needs an account and is paid for with AI credits, after your free requests for the month. You can also upload reference material (PDF/txt/md) or search the web for scholarly sources to include as secondary support.

1. Pick a passage range.
2. Choose your AI provider.
3. Optionally add reference material via upload or "Search the web."
4. Click Analyze — a full chapter can take a few minutes.

Try: Ruth 1, focus question "Where is the pivot of this passage?"

As soon as you pick a passage — before running Analyze — the "Distributional fingerprint", "Biblical DNA", and "Bible Genome" panels below the form load automatically: free, deterministic figures with no AI involved (verb forms, stems, top words, rarest words vs. the whole corpus, and, for Hebrew passages, part-of-speech, clause-role, and rhetorical-relation mixes). Greek passages currently only get the word-rarity half — the app says so plainly rather than showing a broken-looking empty section. "Bible Genome" lists other chapters that share unusually rare vocabulary with this one — click any linked chapter to jump straight to it. These are computed shared-word matches only, not claims of quotation or allusion, and don't yet cross-match Old Testament with New Testament passages or include curated study-Bible cross-references — the panel says so directly.

Self-serve wizards that go straight from a topic or a passage to a finished, ready-to-use piece — no need to build a Research project by hand first. Behind the scenes each one runs the same Theophilus Method (Observe → Analyze → Interpret → Synthesize) automatically, then generates the lesson, sermon, or devotional on top of it — the Method work is available via a "Show the Theophilus Method" toggle if you want to see it, but it's collapsed by default.

1. Pick Topical (type a question or topic and get suggested passages) or Biblical passage (choose the passage yourself, one or several).
2. Optionally set an age group and content level (e.g. children vs. adults, introductory vs. in-depth) so the generated material matches your actual audience.
3. Optionally add a "preparer's focus" — a free-text note on an angle, theme, or prophetic emphasis you specifically want the generated piece to develop.
4. Set duration and tone, then generate.
5. Export to Markdown or PowerPoint (lesson/sermon only — devotionals are markdown-only, they're meant to be short).

How a prepared sermon is shaped. Sermons come out in four movements — Huh? (what this passage is about, opening with an illustration and no technical detail yet — start low, go slow), Aha! (into the text itself: patterns, verb forms, word order, lexis, hapax legomena, and what each one does to the sense), Hmmm… (what it means, where the application begins) and Yes LORD! (one or two concrete steps, or a change of perspective landing in thanksgiving). The conclusion returns to the image the sermon opened with. Exposition takes about a third of the time and application about half — that split is computed from your chosen duration, not left to chance, and changing the number of moves redivides the exposition rather than lengthening it. A table under the sermon shows that arithmetic: each segment's minutes, its word budget at your speaking speed (scripture reading has none — it is read aloud), and where the clock stands after it.

Your stories stay yours. Where the sermon wants a personal story it leaves a dashed Your story here box for you to fill; the app never writes one. If the model writes a first-person anecdote anyway, it is removed after generation and an Anecdote removed flag marks the place. Above every result, Built on your interpretation shows which of Analyze, Interpret and Synthesize it stands on, and the passages with the role each was given.

Points: fewer, and odd where the text allows. You can ask for one to five moves, and one, three or five are preferred — not because odd lists are easier to remember (they aren't, particularly) but because an odd number has a middle: a hinge the sermon can turn on. The passage still wins, though. If it genuinely has four structural units you get four, and the sermon tells you which units forced it, because merging two units to reach a tidier outline would misrepresent the text.

One thrust, several landings. Under "Who is in the room" you can name the groups actually in your congregation — the sermon keeps a single controlling idea and adds a concrete landing for each. No group is ever named aloud. The sermon describes the situation from inside it and lets the person recognise themselves, because saying "if you are divorced" reaches the one it names and exposes them to everyone else in the same breath. Ask it to suggest groups from your passage, then change them freely — you know who is in the room and it does not. Leave it empty for a single general application.

It checks itself before you see it. After drafting, the sermon is measured against the same rubric "Evaluate a Sermon" uses, and every Hebrew or Greek claim it makes is checked against the tagged corpus in code — then it revises itself once. Anything the automatic pass could not resolve is listed at the top of the sermon rather than hidden, so you can see what still needs your judgment.

How a prepared devotional is shaped. It follows the same arc as the Daily Devotional on the home page, with room to breathe: what you take the passage to say, then what the original actually says (one to three observations, depending on the length you chose), why that matters, where it sits in the whole story, worship, and one concrete thing for today. The "turn toward worship" control decides how much of the piece that turn is allowed to take. Unlike the daily entry it also carries a closing prayer, and it goes further into where the passage sits in redemptive history — that is what the extra length is for. Every claim it makes about the Hebrew or Greek is checked against the tagged corpus, and anything that could not be confirmed is shown to you rather than hidden.

Try: Prepare a Devotional, topical mode, topic "trusting God in waiting" — the shortest of the three, good for a first look at how the wizard works.

Before you use any generated lesson, sermon, or devotional — read the actual passage yourself in a Bible you trust, in its full context, and check any historical, linguistic, or doctrinal claim against a reputable commentary or study Bible. See Terms of Service §6 for the full disclaimer.

These three tabs are one connected workspace, grouped together in the sidebar: Research is where you build evidence-linked claims project by project; My Commentary and Research Graph are two different ways of looking back across everything you've built there, organized by passage instead of by project. Each is covered on its own below, followed by a start-to-finish walkthrough connecting all three.

Research

Multi-stage research projects following the Theophilus Method (Observe → Analyze → Interpret → Correlate → Evaluate → Synthesize → Communicate → Apply → Transform) — every claim needs evidence or gets flagged "unevidenced." From a solid research project you can draft a full journal article (Write) or a ministry artifact (sermon outline, lesson plan, devotional — and, from a lesson, a 60-second video script). A Guided Study can become a project too: Make… in Ask & Study carries your own observations, interpretations and applications into its stages.

1. Create a project (passage or topic).
2. Work through each Method stage in order. The row of nine above the stages shows where the project stands — each stage marked yours, AI draft or empty, the next one to do outlined — and takes you to a stage's card when you choose it.
3. Each stage card lists the workflows that work on that stage — Correlate, for instance, lists the Word Study, Literature Review, Scholars Map, Debate Map, Annotated Bibliography and Competing Views — and each name takes you to its section. Under every workflow's heading, the stages it serves are named the other way round. Interpret, Apply and Transform list none: Interpret is the stage drafts themselves, and Apply and Transform belong to the Ministry engines.
4. Beside your claims, the knowledge graph shows one claim at a time with what it is connected to: the verses it is anchored to, the sources it cites, the other claims that share one of its verses, and any word two of its verses share. It is read from the project's own claims as they stand, so a claim you added a moment ago is already in it. Choose Centre the graph here on a claim, or click a claim in the graph, to move the centre. The line under the graph counts the whole project — claims, verses, sources, shared words and links — and says how many words were left out because they appear in only one verse. For a wider walk from a verse or a word, use the Research Graph tab.
5. Once the interpretive stages are solid, use Write or Ministry to generate a finished document.

Try: create a passage project for Ruth 1, then draft its Observe stage.

A project also has several optional tools further down the page, each independent of the others: Competing Views (11 interpretive lenses compared side by side), Literature Review and Gap Finder (over your attached sources), Research Quality Review (coverage across nine fixed analytical dimensions — not a correctness score), Ask the Tutor (Direct, Guided, Coach, or Assessment mode — you can always ask for a direct answer regardless of mode), and AI Research Supervisor (a skeptic's critique against 8 questions, advisory only). Click Collaborators to invite someone by email — they can view and comment, never edit; if they don't have an account yet, they get access automatically the moment they sign up with that address. Projects others have shared with you appear under "Shared with me" on the Research project list. When a project's interpretive stages are solid, scroll down to Write — Journal Article — state a research question, click Propose Outline to review the structure, then Draft Article to generate a full, cited manuscript with the same citation-honesty guard as Literature Review, ready to export as Markdown or DOCX.

My Commentary

Every claim you've approved with "Add to My Commentary" (from any Research project), assembled into one reading view — your own words, verbatim, grouped by the verses they're about. No AI involved: nothing here is generated, only gathered. Works whether or not you're signed in.

1. In Research, open a project's Claims section and click "Add to My Commentary" on any claim that has verse evidence attached.
2. Open My Commentary — the book appears in the Books list.
3. Pick a book and chapter (or click the book in the list) to read every approved claim for that chapter's verses, with a link back to the project it came from.
4. Changed your mind? Click Remove on any entry here (or "Remove from My Commentary" on the claim card in Research) — it leaves this view only; the claim itself stays untouched in its project.

Try: in Research, add a claim to Ruth 1:1 with verse evidence "Ruth-1-1," click "Add to My Commentary," then browse to Ruth 1 in My Commentary.

Research Graph

A map of how your own claims connect to verses, shared vocabulary, and sources — built entirely from data you've already recorded. Deterministic, not AI-generated: every line traces to a real claim, verse, or citation.

1. Click "Refresh graph" (or just visit the tab — it refreshes automatically) to pull in your latest claims.
2. Pick a verse from the picker, or paste a node key (e.g. claim:5) to center the diagram on it.
3. Click any neighbor in the diagram to recenter on it and explore further.

Try: after adding a claim with verse evidence in Research, click the claim's "View in Graph" link to see it centered in the diagram.

Walkthrough: from a claim to your commentary to the graph

The three tabs above are meant to be used together, in this order. A concrete run, start to finish, using the sample Ruth data already on screen:

1. Open Research → Create project → passage, book Ruth, chapter 1 → Create project.
2. Under Claims, type a claim — e.g. "The famine drives Elimelech to Moab" — and click Add claim.
3. On that claim card, click Show Evidence, choose kind verse, type the ref Ruth-1-1, and click Add evidence. The claim's badge changes from "unevidenced" to a coverage score — this step is what lets My Commentary and Research Graph find the claim later; a claim with no verse evidence won't appear in either.
4. On the same card, click Add to My Commentary.
5. Open My Commentary — "Ruth" now appears in the Books list. Click it, or use the book/chapter picker, to see your claim under Ruth 1:1, in your own words, with a link back to the project.
6. Back in Research, click View in Graph on the same claim card — this jumps straight to Research Graph with the claim centered, showing its connection to the verse it's anchored to and any source or shared-vocabulary links. Click a neighbor node to explore outward from there.

Which run made this? Under every AI result in a research project — a stage draft, the research map, competing views, the debate map, the agreement matrix, the gap report, the literature review, the quality review and each Supervisor report — a small line names the run that produced it: the workflow, the provider and model, the tokens it read and wrote, and what you were charged. When you paid with your own key, or the call was one of your free requests, the line says so instead of showing a bare zero. Results made before runs were recorded say that too.

The whole loop takes under a minute once you've done it once — the extra steps (adding evidence, clicking "Add to My Commentary") are what make a claim show up in the other two views; nothing happens automatically from just writing the claim text.

For when you have a topic in mind but not yet a question. Press Explore a topic in the Research tab. Seven steps, listed on the left with a credit estimate beside each AI step; you can leave at any step and resume it from Research. There are three tracks: Article, Book, and Dissertation / PhD-level proposal.

1. Starting point — a short form: your topic, what you want to produce, your field, languages and methods. Free.
2. Shape the topic — the app counts the literature for your wording (free CrossRef metadata), then asks two or three multiple-choice questions to narrow it if it is too broad or widen it if it is too thin. Up to four rounds; it stops asking once the count sits in the range for your track.
3. Candidate questions — three to five research questions, each with the contribution it would make and four scores whose meaning is written beside them. Pick one, edit it, or merge two.
4. Literature scan — searches CrossRef, OpenAlex and Semantic Scholar, then tries alternate wordings (older terms, German and French terms of art, transliterations). Mark each result relevant, not relevant or already answers my question. Every result's Open link goes through My library link if you have set one. One optional deep web run (Exa) is available on your own Exa key, with its cost shown before you start it.
5. Gap check — each candidate gap shows the searches it rests on and the sources that were not searched (ATLA if your library has not configured it, for example). A gap that cannot point to a search is not shown. Gaps are relative to what was searched, not to all scholarship — tick the verification checklist before you rely on one.
6. Feasibility — checks the question against your languages, access and time, asks only what it cannot infer, and ends with go, adjust or rethink.
7. Proposal — an article plan or book proposal, drafted section by section. It may cite only the sources you marked relevant; venues and comparable titles come only from your scan. Run the Supervisor, then Create project: you get an ordinary research project with those sources attached and the proposal open in the writing engine.

Every AI step is paid with credits, or with your own Anthropic, OpenAI or Gemini key if you have saved one in My Profile. The app never calls a topic "original" or "publishable"; it says what it searched and what it found. You can start three explorations a day.

The dissertation track asks for your degree, word limit, years available and (if you have them) your programme's proposal requirements, and goes further at four points:

• Scan — it also searches OpenAlex for dissertations only, and offers ProQuest Dissertations & Theses (through My library link), OATD and TREN as link-outs with the exact terms to search. The app cannot search those three for you, so the gap check lists them as not searched until you have.
• Gap check — two more checks, computed from your results rather than asked of the AI: whether the literature where a question like yours is usually answered (German monograph series, Jewish studies journals, linguistics venues…) is under-sampled, and whether attention to the question peaked decades ago and faded. Either one demotes a gap to suspect and names the reading or searches that would settle it.
• Candidate questions — scored against the full rubric: the meaning of every score is written beside it, a weighted composite is shown, and a 1 on any criterion marks the question not recommended.
• Proposal — problem statement, thesis and research questions, literature review, methodology, contribution, chapter outline, scope and limits, timeline and bibliography, with an appendix giving the gap and its search log (built from step 5, never written by the AI). The Supervisor asks four more questions a committee would ask, and any deficiency it finds must be fixed or acknowledged before Create project.

The Gap Finder (optional, at the top of the gap check, every track) builds a larger corpus around your question — a few hundred records: one large page from each database and the works citing your most-cited approved sources — groups it by shared vocabulary, and scores how much attention each group gets (how many records, how recent, how cited, how many venues). A group close to your question that gets little attention is marked under-studied; a crowded, still-active one is marked crowded and active. A gap the check draws from a group says so, and stays suspect while ATLA was not part of the corpus. Under 60 records, it does not score at all: attention measured in a corpus that small says nothing about a field. Abstracts are read for the grouping and never stored.

ATLA and Research in Ministry are searched only when your institution's library has issued an EBSCO EDS API account for TheophilusBible and we have connected it to your institution's licence. It is the library's account, not yours: ask your librarian, and never send us a personal EBSCO login.

The Topic Explorer's gap check tests each candidate gap against two more of the Gap Finder's checks, on every track. Both are worked out by the app from what it measured; the AI supplies a score or a link, and the app decides whether the check fired.

• Can it be answered yet? (H4) — the AI rates the evidence available for the gap from 1 (essentially none) to 5 (comprehensive primary evidence to hand), using fixed descriptions of each score. At 2 or below the gap is shown as genuine but infeasible: perhaps unanswered because it cannot yet be answered. It stays on the list, since a different method or a comparison may make it answerable.
• Is anyone's argument blocked on it? (H5) — two signs that it matters: other groups of work in the Gap Finder corpus cite into its group (counted from the searches for works citing your approved sources), or a crux on a debate map turns on it. Choose one of your Research projects that has a Debate Map under Compare with a project's debate map before running the check. With neither sign, the gap is marked suspect; whether it is still worth answering is a judgement about its scholarly value, not the check's. If either measure is missing (no corpus, or no debate map chosen), H5 says what was missing and does not count against the gap.

Once a Research project has both Competing Views and a Debate Map, press Build agreement matrix from the debate map's cruxes under Competing Views. For each crux on the debate map, every lens gives its answer, and every pair of lenses is scored from 0 to 3 with one sentence saying why: 0 means their answers are incompatible, 1 different frameworks that are partly reconcilable, 2 compatible with different emphases, 3 substantive agreement.

Pick a crux to see its grid, each lens's answer and the reasons; Overall shows the mean across cruxes. A "·" is a pair that was not scored — the app leaves it empty rather than guessing, and a pair given two different scores is left empty too. It makes two AI calls: the scoring, then the AI Research Supervisor's review. The scores are the model's judgement of the lens readings, not a measurement; read the reasons. Regenerating Competing Views clears the matrix, and changing the debate map's cruxes marks it out of date.

In a Research project, Debate Map lays out a scholarly debate on the project's question as an argument graph: the positions held, the arguments for each, the rebuttals between arguments (one undercuts another's inference, rebuts its conclusion, or outweighs it), and the cruxes — the two to four points, such as a date or the sense of a word, on which the positions actually divide.

1. Attach the sources you are working from, and record any claims you have already established; the map can cite only those, and verses.
2. Optionally type the question, then press Generate Debate Map. It makes two AI calls: the map, then the AI Research Supervisor's review of it.
3. Open a position to see its arguments, each premise with the evidence it cites, and the rebuttals aimed at it.

Every argument on the map cites at least one verse, attached source or claim of yours that the app checked exists. A link that did not resolve, a premise left without evidence, an argument left without a premise, or a scholar who is not among your attached sources is removed and listed under the map, so you can see what was dropped. The check is that the evidence exists, not that it says what the argument uses it for; that is still yours to read. The map is built from your project, not from a survey of the field. Regenerating replaces it.

In a Research project, Make Annotated Bibliography lists every source attached to the project with two short paragraphs under it: what the work argues and on what evidence, then how it bears on your project's question. Choose alphabetical order, or by theme to group related works.

• The entry itself — author, title, publication details — is formatted in SBL style by the same citation pipeline as your exports. The AI never writes a citation.
• The AI is shown only what your library holds for each source: its details and, where you have one, its abstract. A source with no abstract is summarised from its title only, marked Not yet examined in full, and given no relevance note, because its relevance cannot be judged from a title.
• If the AI skips a source, the entry says no annotation was written; nothing is filled in.

It makes one AI call per ten sources, then the AI Research Supervisor reviews the result. Download Markdown or Download DOCX exports it.

Make Scholars Map lists the authors and editors of the sources attached to a Research project — no one else — and asks the AI how they relate: where each stands on the positions of the project's Debate Map (holds, rejects or qualifies), how they relate to each other (agrees with, disputes, responds to, student of), and which schools they might form. It makes two AI calls: the map, then the AI Research Supervisor.

What a link is worth is marked on it:

• shown — the AI quoted the abstract of one of that scholar's own sources, and the quote was found there word for word. Hover to read it. A source you hold without an abstract cannot show anything.
• yours — you recorded it under Add what you have read, with a note saying where. Your links are kept when the map is made again.
• suggested — the AI proposed it but nothing you hold shows it. Treat it as a lead to check, not a fact. School groupings are always suggestions.

In a Research project, type a word's dictionary form (lemma), choose Hebrew, Aramaic or Greek, and press Study this word. The first half is counted from the corpus, with no AI:

• every occurrence, by book and by section of the canon (Torah, Prophets, Writings, Gospels and Acts, and so on);
• for a verb, how often it stands in each stem and tense-form;
• the words it keeps company with — within five words in the same verse — ranked by how much more often they appear together than chance would give (pointwise mutual information). A pair must occur at least three times to be ranked, because rare words otherwise top the list by accident;
• words in the other testament that share its commonest gloss. That pairing is a heuristic: a shared English gloss is not a shared meaning.

Then one AI call sorts a sample of up to 80 verses (all of them, for a rarer word) into senses. A sense may list only verses the AI was shown, and each verse belongs to one sense; anything else is removed and listed. The AI also audits its own senses for four fallacies — reading meaning from a word's root, pouring every sense into one use, reading a later meaning back, and loading a doctrine onto a word — and a check it did not answer shows as not checked, not as passed. The AI Research Supervisor reviews the study.

Everything you've saved from Read, Exegesis, Structures, and Search, organized into folders. Copy or download any item as Markdown, or delete what you no longer need.

1. From Read, Exegesis, Structures, or Search, save a result you want to keep.
2. Optionally create a folder and file the item into it.
3. Open My Library any time to browse, copy, download, or delete saved items.

In a Research project, Thesis keeps every version of the thesis you are arguing and puts each one in front of a critic. Write the thesis, optionally the length you are aiming at, and press Critique this thesis. The critic objects on five axes:

• Precision — are the key terms defined?
• Scope — can it be shown in the length you are aiming at?
• Falsifiability — what evidence would count against it?
• Novelty — how it stands against the positions this project already records (its literature review, if it has one). This compares it with your project, not with all scholarship.
• Evidence reachability — does the evidence you have recorded bear on it?

Each objection is fatal, major or minor, and the critic proposes a sharper statement. Adopt this restatement makes it the next version and critiques that; or write your own next version. Earlier versions are never changed or deleted.

Accept this version is yours to press, but only on the latest version, and only once you have written an answer to each fatal or major objection — the answers are kept with it. The argument outline builds on the accepted thesis. Each critique makes two AI calls: the critic, then the AI Research Supervisor.

Once a project has an accepted Thesis, Build Argument Outline turns it into a tree: the thesis at the top, the claims that support it, the objections it must face (counter), the answers to them (rebuttal), and what it grants (concession). It makes two AI calls: the outline, then the AI Research Supervisor's review.

Each claim carries an evidence obligation, worked out by the app, not the AI:

• satisfied — it links one of your claims whose evidence coverage is 0.60 or more;
• partial — it rests on something that exists (a weaker claim of yours, a verse, an attached source) but nothing that strong;
• open — nothing it cites exists in the project. It says what would satisfy it: a passage to study, a source to find, a word study, or a Competing Views comparison. For a source, Search for it fills in the source search above.

An outline must face its strongest opponent. The AI names which Debate Map positions oppose the thesis; the app picks the strongest of them (the one with the most evidence-linked arguments on the map) and marks the outline invalid if no counter, rebuttal or concession answers it. Without a Debate Map the check cannot run, and the outline says so.

When a project's research is far enough along — its Interpret and Synthesize stages have content — Output Plan plans what it becomes. Choose the kind of piece (conference paper, journal article, monograph, sermon, lesson, blog post, podcast or video), optionally the audience and length, and press Make Output Plan. It makes two AI calls: the plan, then the AI Research Supervisor's review.

The plan gives each section a word budget, the claims and argument-outline steps it carries, how heavily it cites, and the Method stage it draws on. Any claim of yours with coverage of 0.60 or more that no section carries is listed, so leaving good evidence out is a choice you can see. With it come four more parts of the project's dossier:

• Positions — the Debate Map's positions, who holds them among your sources, and the objection each faces. How widely a position is held in the field is not judged: your project's sources cannot show that.
• Open questions — one per crux on the Debate Map, with why it is unresolved and what would settle it.
• Read next — from the Argument Outline's open claims: searches to run (never made-up references), and attached sources you hold only as a title that a claim leans on.
• Hand-off — for a sermon or lesson, the thesis and key claims with their evidence, which the Preaching or Teaching engine under Ministry then reads. Those engines still shape the sermon or lesson.

For a paper or article, Start a document from this plan opens it in Write with the thesis as the research question and the plan's sections as the outline.

Sources found in the Research workspace — search results and the sources attached to a project — carry an Open link to the work itself. Most journal articles, books and dissertations sit behind a subscription your university or seminary library pays for, so on its own that link usually stops at the publisher's paywall.

Add My library link on My Profile and every Open link goes through your library's sign-in instead, the same way it would from the library's own website. You sign in to your library in your browser; the article then opens with your institution's access.

1. Find your library's proxy link. It usually ends in url= — for example https://ezproxy.your-library.edu/login?url= or https://go.openathens.net/redirector/your-library.edu?url=. Your library's "off-campus access" page lists it, or a librarian can tell you.
2. Paste it into My Profile → My library link and press Save.
3. Use "open a sample article through your library" to check it: you should see your library's sign-in page, then the article.

TheophilusBible never asks for or stores your library username or password. Library databases license their content to your institution for your own reading, and none of them allow another service to sign in as you — so the link sends you to your library, and your library does the signing in.

Manage your AI credits and opt in to research memory.

• AI credits — see your balance, buy a one-time credit top-up, or set up auto top-up.
• Research memory — off by default; when enabled, facts you save (via "Save to memory" buttons around the app, or typed directly here) are available to the AI Tutor for continuity across projects. Always yours to edit or delete.
• My library link — your library's proxy link, so the Open links on research sources go through your institution's sign-in. See "Opening sources through your library" above.

The interface can display in English plus fifteen other languages — Indonesian, Simplified Chinese, Burmese, Thai, Tamil, Tagalog, Hebrew, Greek, Japanese, Korean, Spanish, German, Italian, French and Dutch. Pick one from the language switcher in the sidebar. English is the only one that has been through human review; every other language is in beta — machine-translated and not yet checked by a fluent theological reader, so treat the wording as approximate and the original as authoritative. Separately, most AI-powered tools let you choose an output language for the generated content itself (which can differ from your interface language) — look for a language selector near that tool's controls.

Neither of these is an "account link" — Logos doesn't offer one, and only one of the two Zotero paths below crosses the network at all. Read the fine print before you rely on either.

Zotero — file import (works everywhere, including this hosted site). The reliable path. In Zotero: File → Export Library…, format BibTeX or RIS. In a research project's Sources section, upload the resulting .bib/.ris file.

Zotero — live "Pull from Zotero" button. Only works when you're running TheophilusBible locally (the desktop app, or theophilus serve on your own machine) with the Zotero desktop app also running on that same computer, with Zotero's Settings → Advanced → "Allow other applications on this computer to communicate with Zotero" turned on. On the hosted site this button will always fail — "localhost" means Render's own server, never your computer, so no amount of getting Zotero running on your end fixes it. Use the file import above instead.

Logos. Local-only, macOS-only, and there's no cloud account to connect — the integration reads Logos's own on-disk library/notes databases directly and sends deep-link URLs to open a passage in the Logos app. This only works running TheophilusBible locally on the same Mac that has Logos installed; it cannot work on the hosted site at all, on any platform. If that's your setup: open a passage-based research project, and Connectors will show logos: available ✓ along with an "Open [passage] in Logos" button and a catalog search box. If it instead shows "unavailable," the reason listed there is the actual cause (not on a Mac, or Logos has never been run). Importing your own Logos notes as a source is possible today only from the command line (theophilus logos open / theophilus logos catalog) — there's no button for that specific action yet.

Accent color, font pairing, text size, and light/dark mode — all client-side, remembered across visits.

Two reading themes sit alongside light and dark: Sepia, a warm paper tone for long reading, and Midnight, a darker and lower-contrast variant of dark mode. They change the page's colours only — nothing about the text, the analysis or what is saved.

1. Click Appearance in the sidebar.
2. Pick a preset or customize each option.
3. Changes apply instantly — no save button needed.

You don't need an account to read. The Library, Search, Structures, Diagram, commentary and the daily devotional are open to everyone. Sign in (Google, GitHub, Microsoft, or an email and password) to use the AI tools, which are paid for with credits, and to keep your own projects, uploads and flashcards.

Signed in somewhere you shouldn't have stayed signed in? My Profile has Sign out everywhere else: it ends your session on every other browser and device and keeps this one. Resetting your password does the same.

Everything above lives inside this app, which means a link to it opens the app rather than the thing you wanted to show someone. A few pages exist outside it as plain web pages — they open in any browser, need no account, and read properly when pasted into a message, an email, or a post.

• Today's devotional, and every past one at its own permanent address (browse the archive) — the dated link is the one to share, since it keeps pointing at that entry.
• Samples — one example of each kind of work, for showing a colleague what the platform produces.
• The front page, which explains the platform to someone who has never heard of it.

These are also the pages search engines can read. The app itself is invisible to them, which is why the public pages exist.

Feature request, feedback, or a bug — click Contact us in the sidebar. No account needed, and we read every message.